Abstract
404
Objectives We investigate the accuracy of a new method which reconstructs activity distribution inside a region of interest (ROI) by iteratively correcting for both spill-in and spill-out effects.
Methods Assuming that boundaries of the investigated organ/tumor are known (e.g. from CT scan), we divide the reconstructed image into two parts: ROI and background. To correct for spill-out effect, 3D recovery matrix is calculated via projection and reconstruction of a template - a numeric phantom whose values are set to unity inside ROI and to zero outside of it. Our algorithm iteratively repeats three steps: (a) analytical projection of both ROI and background (projection step); (b) reconstruction of ROI-image only by incorporating background projections into the forward step of OSEM algorithm (correction for spill-in); (c) multiplication of values of the image inside a ROI by corresponding elements of a recovery matrix (correction for spill-out). Attenuation (AC), scatter (SC) corrections and resolution recovery (RR) are implemented into OSEM. We tested our algorithm using phantom data acquired on SPECT-CT GE Infinia-Hawkeye with 128x128 matrix and 120 views. The ratio between Tc-99m concentrations in heart insert (ROI) and thorax (background) was close to 6:1.
Results Initial image reconstructed with RR+AC+SC contained only 70% of true activity inside ROI. After 4 iterations of the new method, this percentage was improved to 94-99% depending on the SC tactic.
Conclusions The accuracy of reconstructed activity inside CT-based ROI can be substantially enhanced without any assumptions regarding background concentration. The contribution of background to the spill-in effect can be iteratively estimated.
- © 2009 by Society of Nuclear Medicine